How to Set Up an Airtable Field Agent

And why you might use field agents.

Line drawing of a cute robot with the Tinkering With Ideas logo

A few weeks ago, I wrote about how I use Airtable for content. I specifically touched on AI, and how I hadn't found any use cases for AI in Airtable.

I take it back! The use cases I'd uncovered – like asking Airtable's AI assistant about my data or using it to build entire bases – weren't that interesting to me. But AI field agents are incredibly useful.

Field agents can automatically populate data into a field based on a prompt. Airtable had lots of ways to populate data before, but it was rigid. Like "If this thing happens, update this other field with a specific piece of data." Field agents have a lot more flexibility.

I'm using field agents to automatically select categories for my data. Full disclosure: I was doing this through Zapier before (with ChatGPT). But now, I think it makes more sense to do it directly in Airtable.

Here's how it works and why you might want a field agent in Airtable [affiliate link].

Step 1: Add a new field to your table

First, you'll create a new field by clicking on the (+) button.

Adding a new field in Airtable
Adding a new field in Airtable

From the field types, select "Build a field agent."

Select "Build a field agent"
Select "Build a field agent" in Airtable

Step 2: Configure your field agent

You'll give your field a name, just like you would any other field in Airtable.

Naming the new field
Naming the new field in Airtable

Next, you'll select a Field Type. For this example, I'm going to categorize my data based on another field in Airtable. I'll use a "Single Select" field type.

For context, this is a table that contains a bunch of my previously published LinkedIn posts. I want to re-publish some of the posts on other platforms (like Threads or Bluesky) – but only if the original post is about certain topics. To find those, I need to categorize the posts.

Selecting the field type in Airtable
Selecting the field type

You'll have some additional settings, such as "Enable Internet Search" and "Run Automatically." I turned both of these on.

You can also select the Model to use, and I left it at the Default.

Configuring the field settings
Configuring the field settings

Step 3: Add your instructions

Next, you'll type your instructions for the field agent, much like you would use in ChatGPT. If you're looking to categorize data with a field agent, here's the prompt I used:

Categorize the following text into one of the defined categories below. Only return a single option.

Here's a brief description of how they should be categorized:

Work Culture: Toxic work environments, bad jobs, bad bosses
Automation and AI: Tools and tips about different products or use cases
Remote Work: Working remotely or working from home
Freelancing or Solo Work: Working as a freelance writer or solopreneur
Other: Doesn't fit into one of the other categories

After my prompt, I clicked the "Insert Field" button.

Inserting data from another field in Airtable
Inserting data from another field

I inserted data from the same table. The "text" field is my original LinkedIn post. That way, the field agent knows what data it is categorizing.

Selecting my data from the table
Selecting my data from the table

Step 4: Add your single-select options

Since my field type is "Single select," I had to add my single select options. These match the instructions I used in my prompt.

Like any single-select field in Airtable, you can color code the options and change the order in which they are listed.

Adding single-select field options
Adding single-select field options

Once you're done, click the "Save" button at the bottom.

Saving the AI field agent
Saving the AI field agent

Step 5: Let the field agent run

If checked, your field agent will run automatically. This table had thousands of records, so it took a few minutes to populate the data the first time. Going forward, any new records will have the category automatically added.

I can still make changes. If the Category assigned wasn't correct, I can change it by selecting one of the other options.

The categorized data in Airtable, courtesy of the field agent
The categorized data in Airtable, courtesy of the field agent

I should note that using Airtable AI is based on a credits system that I don't fully understand. Airtable's pricing page isn't clear either, other than that different actions take different credits.

I'm on a team plan of Airtable with two users (myself and my virtual assistant). I have 30,000 AI credits per month. Setting up this field and testing it a few times used up 4,667 credits.

I expect that setting up a field and retroactively categorizing a bunch of data could burn through a lot of credits. But running a field agent on new records might not be too bad, depending on how you use Airtable.

I should add that the AI credits are included in my Airtable plan. I'm not paying extra for them.

Automatically Repurpose Social Posts Using ChatGPT
Automatically tweak your posts before publishing them on another social platform.

More resources:

CTA Image

Download my free resource for more tips.

Download